Genomic Biomarker for Prostate Cancer Focal Therapy: Post Hoc Assessment of a Phase II Clinical Trial

Abstract

PURPOSE

A biomarker to help predict outcomes after prostate cancer (PCa) focal therapy would be of considerable interest. We sought to assess the association between treatment failure after focal therapy and the Decipher score, a tumor-based genomic classifier (GC).

MATERIALS AND METHODS

We performed a post hoc analysis of a single-center phase II trial (ClinicalTrials.gov identifier: NCT03503643) in which patients with unilateral grade group (GG) 2-4 PCa (n = 108) underwent hemigland cryoablation of the prostate (2017-2021; n = 108). Pretreatment biopsy tissue was subjected to transcriptomic profiling to generate GC scores. The primary outcome was the association between GC-low (<0.45) versus GC-high (≥0.45) and in-field recurrence (GG ≥2) on magnetic resonance imaging–guided biopsy 6 months post-treatment, evaluated using multivariable logistic regression.

RESULTS

In the GC-high group (n = 37), treatment failure occurred in 17 patients (46%). In the GC-low group (n = 71), treatment failure occurred in 15 patients (21%). These differences were statistically significant (odds ratio [OR], 2.61 [95% CI, 1.05 to 6.51]; P = .04). Differences at 18 months were also significant (76% v 44%; OR, 3.58 [95% CI, 1.37 to 9.36], P = .009).

CONCLUSION

In patients with PCa otherwise suitable for management with focal therapy, a high GC score (≥0.45) was independently associated with treatment failure. A GC score derived from diagnostic biopsy can be used to help predict focal therapy outcomes.

INTRODUCTION

Focal therapy of prostate cancer (PCa) is of increasing interest because it may provide cancer cure while preserving quality of life.¹ The favorable outcomes of partial gland ablation versus whole-organ treatment have been shown in retrospective and prospective studies.^2-4 Largely a consideration for localized intermediate-risk PCa, focal therapy has been shown to achieve near-term oncologic results comparable with surgery or radiation, with fewer urinary, sexual, and bowel complications.^2,4

CONTEXT

• Key Objective

• Can a genomic classifier derived from diagnostic biopsy tissue identify patients with localized prostate cancer who are more likely to experience early treatment failure after hemigland focal cryotherapy?

• Knowledge Generated

• Among 108 patients treated with hemigland cryotherapy, treatment failure occurred in 46% of those with high Decipher scores (≥0.45) versus 21% with low scores (<0.45) at 6 months. This association persisted at 18 months, with failure rates of 76% versus 44%, respectively.

• Relevance

• The Decipher score, derived from pretreatment biopsy tissue, may help identify patients at higher risk of failing focal therapy. This can guide personalized treatment selection between focal therapy and definitive approaches such as prostatectomy or radiation.

However, within 2 years of treatment, up to one third of patients undergoing focal therapy experience disease recurrence.^5,6 Issues regarding tumor targeting and patient selection remain unsolved. Magnetic resonance imaging (MRI) guidance has improved treatment targeting, but criteria for patient selection are still evolving.⁷ Among the clinical variables proposed as selection criteria are those related to tumor grade, size, and location.

Genomic markers, which can identify tumors as inherently aggressive and treatment-resistant, remain an area of exploration. The Decipher prostate genomic classifier (GC) has been shown to provide independent prognostic information in men with PCa undergoing radical prostatectomy and radiation therapy.^8,9 This GC score, which is obtained from paraffin-preserved tissue samples, measures the expression of 22 transcripts associated with aggressive PCa. In this prospective trial, we study the outcomes of PCa focal therapy in relation to baseline GC scores.

MATERIALS AND METHODS

Patients and Trial Description

This study included patients enrolled in a prospective, open-label, single-center trial (ClinicalTrials.gov identifier: NCT03503643).⁵ For this work, they met the following criteria: Unilateral ≥ grade group (GG) 2 PCa diagnosed by MRI-guided prostate biopsy (MRGB; 2017-2021), no previous prostate PCa treatments, and life expectancy ≥10 years. Per-protocol MRGB was performed in all patients at 6 months (sampling the treated lobe) and, in those without significant PCa recurrence, again at 18 months (sampling bilaterally). All patients provided informed consent for trial participation, and details of the focal cryotherapy and MRGB are reported elsewhere.⁵

Although the original trial included patients with GG1 PCa, we focused on patients with higher-grade disease since the preferred management approach for patients with GG1 is surveillance.^10,11 Initial surveillance, as opposed to active treatment, can help patients avoid or delay the adverse effect of active treatment.¹² While this approach is generally considered safe for patients with GG1 PCa, initial treatment is typically recommended for patients with ≥GG2 disease.

RNA Profiling

After institutional review board approval (IRB No. 23-001539), RNA profiling was performed retrospectively on archived biopsy tissue after the trial's completion. Pathologic review of baseline tumor biopsies was performed (A.E.S.) to facilitate microdissection of five 5-μm thick formalin-fixed, paraffin-embedded tissue sections. All RNA extraction, cDNA amplification, oligonucleotide microarray hybridization, and quality control to generate GC scores were conducted in a Clinical Laboratory Improvement Amendments–certified laboratory (Veracyte Inc, San Diego, CA) as described in previous works.^13,14

Signatures

Previously validated gene signatures were the exposures of interest. These were all chosen a priori. The commercially available Decipher prostate GC was chosen as the primary exposure of interest, whereas all others were exploratory. The GC was previously validated to predict risk of metastatic recurrence after radical prostatectomy and has subsequently been assessed in other clinical cohorts as prognostic of outcomes following other management approaches.^8,9 This GC score is based on the expression of 22 genes and ranges from 0 to 1, with higher values suggesting more aggressive PCa. To optimize the clinical applicability of this study, we assessed the GC using a typical cutoff of 0.45: low-risk GC (GC-low) <0.45 and intermediate- or high-risk (GC-high) ≥0.45. This previously determined cutoff is based on the commercially used cutoffs from previous validation multicohort studies of nearly 1,000 patients.¹⁵ The GC was also assessed as a continuous variable per 0.1-unit increase as a secondary analysis. Details of exploratory gene signatures are provided in the Data Supplement (Table S1).^13,14,16-22 All these were selected a priori for this analysis and were assessed as categorical and continuous variables.

Outcomes

The primary outcome was the presence of any ≥GG2 PCa on per-protocol MRI-guided 6-month prostate biopsy (failure). The secondary outcome was ≥GG2 PCa on either the 6- or 18-month biopsy (failure). Salvage treatment after cryotherapy was recorded within the trial period. Since follow-up was short and many patients might have received treatment outside of the trial institution, salvage treatment was classified as “Yes” or “No/Unknown.”

Statistical Analyses

We followed EQUATOR guidelines for tumor marker prognostic studies (REMARK; Appendix).²³ To provide a parsimonious multivariable model for outcome assessment, we used data from the previous trial and pathologic variables known to be associated with aggressive PCa to select covariates a priori.^5,24 We included pretreatment serum prostate-specific antigen density (PSAD) using prostate volume on MRI (Continuous per +0.1), GG (Categorical; 2 v 3-4), and Prostate Imaging Reporting & Data System (PI-RADS) score on MRI (Categorical; 1-3 v 4-5). Each of these covariates were part of the prospective trial data collection.

The primary and secondary outcomes were assessed in multivariable logistic regressions adjusting for all covariates and one of the gene signatures to generate adjusted odds ratios and 95% CIs. P values < .05 were considered statistically significant. All P values were two-sided, and all analyses were conducted using R version 4.3.2 (Vienna, Austria).

RESULTS

Patient and Tumor Characteristics

Of the 131 patients enrolled in the trial with baseline GG ≥2 PCa, 108 (82%) had baseline biopsy tissue available for profiling. The median PSAD was 0.16. Most patients had GG2 PCa (n = 70, 65%; Table 1). The PI-RADS score was most often 4 or 5 (n = 89, 82%). The median time between prebiopsy MRI and cryotherapy was 3.9 months. The median GC score was 0.34, and 71 patients (66%) had a GC score of <0.45 (GC-low). Patients with GC ≥ 0.45 (GC-high) more often had GG3 or 4 PCa (n = 15, 41%) compared with those with GC-low (n = 23, 32%; Data Supplement, Table S2). Patients with GC-high were more likely to have PI-RADS four or 5 (n = 33; 89%) compared with those with GC-low (n = 56, 79%). Characteristics based on an a priori list of exploratory signatures are provided in the Data Supplement (Table S3).

TABLE 1.

Patient Characteristics

Characteristic	No. (%)/Median IQR
Total	108 (100)
Age, years	68 (63-74)
PSA, ng/mL	6.7 (4.7-10.4)
PSA density	0.16 (0.11-0.24)
Grade group
2	70 (65)
3-4	38 (35)
PI-RADS
1-3	19 (18)
4-5	89 (82)
Months between MRI and cryotherapy	4.0 (2.8-7.1)
GC	0.34 (0.20-0.50)
GC-low (<0.45)	71 (66)
GC-high (≥0.45)	37 (34)

Abbreviations: GC, genomic classifier; MRI, magnetic resonance imaging; PI-RADS, Prostate Imaging Reporting and Data System; PSA, prostate-specific antigen.

High Versus Low GC

In the 6-month biopsy, 32 of 108 patients (30%) had ≥GG2 PCa on biopsy (failure). Patients with GC-high were more likely to have GG ≥2 PCa in the treatment field (46%) compared with those with GC-low (21%; odds ratio [OR], 2.61 [95% CI, 1.05 to 6.51], P = .039; Table 2, Figs 1 and 2, and Data Supplement, Tables S4 and S5). A total of 54 patients (55%) had ≥GG2 on either their 6- or 18-month biopsy. GC-high was also associated with GG ≥2 PCa in the 6- or 18-month biopsy (76% v 44%; OR, 3.58 [95% CI, 1.37 to 9.36], P = .009).

TABLE 2.

Multivariable Logistic Regression Results for Focal Therapy Failure in 6- and 18-Month Biopsies With Exploratory Signatures

Covariate	6-Month Biopsy		18-Month Biopsy
	OR (95% CI)	P	OR (95% CI)	P
PSC: non-LD v LD	2.87 (1.01 to 8.11)	.047	2.38 (0.99 to 5.68)	.052
PAM50: Luminal B v Luminal A	1.24 (0.36 to 4.26)	.7	0.88 (0.28 to 2.77)	.8
PAM50: basal v Luminal A	1.45 (0.42 to 5.02)	.5	0.96 (0.30 to 3.08)	.9
RB loss v no loss	0.57 (0.13 to 2.44)	.5	0.68 (0.18 to 2.62)	.6
PTEN loss/indeterminate v wild-type	1.75 (0.63 to 4.91)	.3	2.09 (0.71 to 6.16)	.181
P53-mutant v wild-type	3.02 (0.73 to 12.62)	.129	2.56 (0.48 to 13.73)	.3
ERG-positive v ERG-negative	1.70 (0.69 to 4.17)	.2	2.02 (0.83 to 4.89)	.119
Low AR activity v higher	4.77 (0.88 to 25.86)	.070	5.78 (0.60 to 55.85)	.130
HR-deficient v HR-intact	0.59 (0.17 to 2.07)	.4	0.70 (0.21 to 2.28)	.5
FOLH1: Q2-Q3 v Q1	1.18 (0.39 to 3.56)	.7	0.86 (0.30 to 2.44)	.782
FOLH1: Q4 v Q1	1.13 (0.32 to 4.00)	.8	1.09 (0.33 to 3.67)	.9

NOTE. A total of 108 patients had data available for 6-month biopsies with 32 (30%) failures. A total of 99 patients had data available for 18-month biopsies with 54 (55%) failures. Citations for each signature can be found in the Data Supplement (Table S1). All rows present the regression results for the given signature adjusted for grade group, prostate-specific antigen density, and Prostate Imaging Reporting and Data System. Q1-4 are quartiles based with Q1 being the highest value. Bolding indicates statistical significance.

Abbreviations: AR, androgen receptor; ERG, ETS related gene; HR, homologous recombination; LD, luminal differentiated; OR, odds ratio; PSC, prostate subtype classifier; Q1, quartile 1; Q2, quartile 2; Q3, quartile 3; Q4, quartile 4; RB, retinoblastoma gene.

FIG 1.

Focal therapy failure percentages at 6 and 18 months based on a GC. Upper error bars are 95% CIs based on Pearson's chi-squared distribution. Treatment failure was defined as any grade group ≥2 cancer on post-treatment biopsy. The cutoff score for GC-high and GC-low was based on a previously defined cutoff for a low-risk GC of <0.45.³¹ GC, genomic classifier.

FIG 2.

Multivariable logistic regression results for failure at 6- and 18-month biopsies. A total of 108 patients had data available for 6-month biopsies with 32 (30%) failures. A total of 99 patients had data available for 18-month biopsies with 54 (55%) failures. Treatment failure was defined as any grade group ≥2 cancer on post-treatment biopsy. The cutoff score for GC-high and GC-low was based on a previously defined cutoff for a low-risk GC of <0.45.³¹ GC, genomic classifier; GG, grade group; PI-RADS, Prostate Imaging Reporting and Data System; PSAD, prostate-specific antigen density.

Exploratory Analyses

As a continuous variable, higher GC scores were associated with increased odds of focal therapy failure although this association was not statistically significant (Data Supplement, Table S6). Of the exploratory genomic signatures, only the Prostate Subtyping Classifier (PSC) was associated with treatment failure at 6 months (Table 2 and Data Supplement, Table S7).¹⁴ Nonluminal differentiated (non-LD) tumors (n = 70) were associated with higher odds of failure at 6 months (37%) compared with LD tumors (n = 38; 16%; OR, 2.87 [95% CI, 1.01 to 8.11], P = .047). The 18-month assessment showed a similar, albeit nonsignificant trend (OR, 2.38 [95% CI, 0.99 to 5.68], P = .052).

Of the 29 patients with GC-low who failed by 18 months, eight (28%) received a second cryotherapy, one underwent a second cryotherapy followed by prostatectomy, and two underwent prostatectomy alone. Of the 25 patients with GC-high who failed by 18 months, one underwent brachytherapy, three (12%) underwent a second cryotherapy, and one underwent prostatectomy.

DISCUSSION

First described some 20 years ago,²⁵ the use of focal therapy is now increasing, and it has recently appeared as an investigational treatment option in some guidelines.^10,11,26 However, failure rates—defined as persistence or recurrence of ≥GG2 tumor—remain a major concern after focal therapy.^5,6 Therefore, there is a need to characterize biomarkers that can identify potential candidates for focal therapy who are at a high risk of failure and might instead benefit from up-front prostatectomy or radiotherapy.⁷ In this post hoc assessment of a prospective trial, we found that patients with a high GC score based on pretreatment tissue were more likely to experience treatment failure than those with low GC scores.

Treatment failure after focal therapy for PCa can lead to higher health care costs, need for additional treatments, and a risk of cancer progression. There is no universal definition of focal therapy success or failure, and follow-up surveillance can also vary in the use and timing of imaging and biopsies.²⁷ Monitoring after focal therapy may include MRI 6-12 months after treatment. Biopsies are then performed when there is a concerning lesion on MRI (for cause)²⁸ or routinely at 6-12 months after treatment (per protocol) as in this study. This method reveals a stringent and higher rate of treatment failure than when only clinical end points are used.²⁸

In this study, patients with ≥GG2 PCa experienced treatment failure at rates of 30% at 6 months and 55% by 18 months. While an optimal failure rate is not defined for focal therapy, patients with GC-high recurred at rates of 46% at 6 months and 76% at 18 months compared with 21% and 44%, respectively, for GC-low. The current methods for identifying the optimal patients for focal therapy are based on clinicopathologic features.²⁹ These factors include prostate volume, tumor volume, grade, extent, prostate-specific antigen (PSA) levels, and PSA kinetics. However, while many consensus statements exist, there are no universally accepted criteria for focal therapy candidacy.²⁹ Even after accounting for common clinicopathologic risk factors for PCa aggressiveness such as PSA density, grade, and mpMRI PI-RADS score,²⁴ GC-high was associated with an increased odds of treatment failure compared with GC-low. Therefore, the GC Decipher score could help identify patients most likely to fail focal therapy and, in some cases, support consideration of radical prostatectomy or radiation. Importantly, the GC scores were derived retrospectively, and thus, this study does not reflect how GC scores might influence practice.

Our exploratory analysis identified a previously validated signature that subtypes prostate tumors based on basal versus luminal cell of origin and was associated with the primary outcome.¹⁴ This signature distinguishes tumors based on basal versus luminal origin, with the luminal LD subtype characterized by slow cell division and high responsiveness to androgen deprivation therapy. Patients with non-LD tumors had higher odds of focal therapy failure at 6-month biopsy. LD tumors are defined by high androgen receptor output and low cancer cell replication rates and have been associated with longer progression-free survival after radical prostatectomy and salvage radiation with hormone therapy.^14,30 They predominate in low-grade, GG1, disease. Our findings suggest that patients with LD tumors may be optimal candidates for management strategies that avoid definitive radiation or surgery, such as focal therapy or surveillance.

Limitations of this study include possible selection bias in the cohort. A total of 23 of 131 (18%) patients with ≥GG2 from the original clinical trial did not have baseline tissue available for molecular profiling. In addition, the primary outcome for this study was treatment failure at 6 months, representing near-term follow-up. This does not account for longer-term follow-up, including death from PCa or time to metastatic (incurable) recurrence. This study was conducted within a clinical trial at a tertiary care center in an urban setting, which may limit generalizability. Notably, these patients participated in a clinical trial with standardized data capture at a tertiary care center under the care of experienced investigators from urology, radiology, and pathology.⁵ Thus, treatments and assessments of recurrence were likely optimized. Thus, ongoing work is needed to provide validation to substantiate the clinical use of the GC to inform decisions about future focal. Finally, salvage treatment follow-up was limited to the short-term follow-up of the clinical trial and the fact that many patients likely received treatments outside of the trial institution.

In conclusion, while focal therapy for PCa may offer oncologic benefits with fewer adverse effects, the rate of treatment failure remains appreciable. A GC based on biopsy tissue identified patients at an increased risk of treatment failure, beyond clinicopathologic variables. Thus, GC may be useful in predicting focal therapy outcomes and counseling patients on treatment failure risk.

DATA SHARING STATEMENT

A data sharing statement provided by the authors is available with this article at DOI https://doi.org/10.1200/PO-25-00535.

AUTHOR CONTRIBUTIONS

Conception and design: Adam B. Weiner, Mamdouh Aker, Michelle Cardenas, Samantha Gonzalez, Elai Davicioni, Wayne G. Brisbane, Leonard S. Marks

Financial support: Adam B. Weiner, Leonard S. Marks

Administrative support: Adam B. Weiner, Elai Davicioni, Leonard S. Marks

Provision of study materials or patients: Adam B. Weiner, Elai Davicioni, Leonard S. Marks

Collection and assembly of data: Adam B. Weiner, Mamdouh Aker, Michelle Cardenas, Samantha Gonzalez, Eileen Kelly, Elai Davicioni, Wayne G. Brisbane, Leonard S. Marks

Data analysis and interpretation: All authors

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/po/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).